The Codex Faenza manuscript 1 is an unique source of late medieval ornaments. The manuscript shows figurations over plain chant lines and ornamentation of existing compositions. Ensemble Currentes has been using this as a source of ornaments for existing pieces. Slower note values and basic structures can be replaced to create a richer surface. Figures can be memorized and spontaneously inserted. These figures become vehicles to stand still or move steps upwards or downwards by different intervals. The first note of an ornament is not at all always a consonance to the plain chant line, but it is circling around a consonance.

I turned to Codex Faenza to collect ornaments for my own use. The first Kyrie has mostly one plainchant tone per measure in the modern notation of the Plamenac edition. 2 Piece number 28 has rapid ornamentation in tempus imperfectum around "La dolce sere" by Bartolino da Padova.

This patch defines criteria for ornamentation.

  • The duration of an ornamented note is between 50 and 5500 milliseconds. This will open for extremely rapid, as well as slowly augmented ornaments. The curve shapes are randomized.
  • With other criteria met, there is a 50-99 percent probability of ornamentation.
  • The possible interval scaling for unison intervals is between .1 and .9, meaning the intervals will be diminished to various degress.

Curve shapes can emulate a musician shaping the tendencies through a piece of music. This ornamentation method is not sensitive to mode, and does not match ornamentation figures to steps in a scale. This would be necessary to create stylistic ornamentations. Microtonality and disregard for modal structure is not part of this late medieval style. Ornament speeds are potentially not quantified to common subdivisions of a piece, allowing intricate polyrhythms through stretching. This ornamentation machine creates distortion of a style, by intended features and by what it is lacking. Distortion was central to my involvement in the "Wheels within wheels" project. When the machine fails, the composer can intervene and make choices.  Let us hear one of many possible Dufay ornamentations.

It is ornamented by this process in different levels of fractalization.


Using the whole Codex Faenza, the selection of figures would be much larger, and the most common figures would have a higher probability faced with random selection. These ornaments became inputs of an ornamentation machine for the OM-Ruben library, named 'r-ornamenter'. Ornamentation should replace simple patterns with richer figurations. This took the form of a search function with input parameters.

  • What is the minimum and maximum length of an ornamented note? A curve shape defines the trajectory of this treshold.
  • What is the minimum and maximum percent probability of ornamentation to happen at a given point? This is also controlled by a curve shape.
  • What is the interval from one note to the next, which of our input ornaments are matching the interval, which are closest to the duration? If there are many results, a random selection will happen. If there are no ornaments for this exact interval, pitches will be microtonally stretched from the closest match. The duration of an ornament will be stretched or compressed to the note duration of the music to be ornamented.
  • If the interval is unison, should we microtonally stretch the interval range? The stretching range is controlled by a curve shape.

The following demo patch shows Guillaume Dufays motet "Nuper rosarum flores" ornamented by figures from the mentioned parts of Codex Faenza. Each rendering of this patch will create a differently ornamented version of Dufay. It is possible to experiment by replacing the piece and the repertoire of ornaments.

Allowing very short notes to be ornamented creates ornamentation at rapid, almost unplayable speeds. This may bring associations to other musical traditions of the world. In order to avoid simplifications, I will not name a particular one.

Replacing long notes with typical ornaments was based on what the Ensemble Currentes was doing. I did also try a fractal approach. Patterns in nature expand from simple patterns to infinite detail. In computer graphics, a mountain can be drawn by endlessly subdividing lines within a triangle. 3

A melodic pattern is analyzed, and we assume that a turning point is by a change between ascending and descending movement. A series of intervals between two turning points is a segment considered a potential ornament. At the middle between two turning points of the melody, a random ornamentation can be inserted. You can choose whether it should be original, retrograde, inversion or retrograde inversion, and whether intervals should be microtonally scaled by floating numbers. This process generates a form of self similarity, as a melody is ornamented with fragments of itself. This continues recursively, the same process is repeated on the result. It does not take many iterations to expand some hundred notes to thousands or millions of notes.

This fractal self-similarity was implemented in the OM-Ruben library through a function called 'r-chord-to-granular-pitch-inflection'. In a demo patch I show an example of using this for granular synthesis with Csound. It was possible to create a dense cloud of notes deduced from randomized melodies or series of figurations from Codex Faenza.

This is the beginning of the diminution line for "Soto l'imperio" from Codex Faenza.

This is of course a grave simplification of how ornamentation works in a musical repertoire. When it is all ups and down, we are missing all the curlings and fluctuations around a pitch which we will find a lot of in the Codex Faenza manuscript. We have zoomed to far in. It is creating a world of self-similarity and repetition, while smoothening out stylistic characteristics of the original. Finding more sophisticated principles of melody segmentation would be a possible way to progress.